The November 2022 release of .NET 7 offered another opportunity to examine runtime performance. The useful question for a product team is narrower than a benchmark headline: does this application’s important workload improve?
Choose a representative journey
Use a workflow such as searching a customer register and opening a result. Include realistic data volume and a mix of common and uncommon queries. A tiny development database can conceal the cost that dominates production.
Capture throughput, latency percentiles, memory use and error rate. Keep concurrency and infrastructure constant between runs. Record the warm-up procedure and distinguish first-request behaviour from steady-state behaviour.
Change one variable
Upgrade the runtime separately from a query rewrite or a cache change. Otherwise, a faster result tells you little about which change mattered. Retain a baseline build so another developer can repeat the comparison.
If the database dominates the request, a runtime improvement may have little visible impact. That is still useful evidence: it directs the next investigation toward indexes, query shape or unnecessary round trips.
Finish with a short decision record containing the workload, environment, measurements and remaining uncertainty. Do not extrapolate one endpoint’s result into a claim about the whole product. For work today, repeat the exercise on supported runtime versions.